Sample Augmentation Method for Side-Scan Sonar Underwater Target Images Based on CBL-sinGAN

Round 1

Reviewer 1 Report

Comments for author File: Comments.pdf

Author Response

Dear Editor and Reviewers,
Thanks very much for taking your time to review this manuscript. l really appreciate allyour comments and suggestions! Please find my itemized responses in below and myrevisions/corrections in the re-submitted files. Thanks again!

Q1

GBL-sinGAN doesn't appear in this article, I guess you mean what CBL means, CBL means the two modules that I added to the original network, the CBAM module and the L1loss loss function.

Q2

The previous work in the introduction has elaborated on the effect of style migration for networks such as cycleGAN with a small amount of data, which has been elaborated in the discussion section of Section 4.

Q3

It is true that we have repeated the description of the network structure in both the introduction and the methodology, and the description of the method section has been modified.

Q4

Section 2.1 is my description of the overall network architecture, which is based on the original paper sinGAN, but there are three adjustments in the structure, and the network structure is different from sinGAN, so it is not plagiarism. Of course, my terminology and descriptions have been changed accordingly, and the language has been reorganized and streamlined according to your suggestions.

Q5

The integration process of the L1 loss function has actually been elaborated in Section 2.1, and I have simplified Section 2.1 by substituting the description into Section 2.3 based on the fourth question

Q6

The parameters at the propagation boundary are generated according to the function notes, that is, the position of the boundary parameters is different for each generated image. This is how the sinGAN function is generated.

Q7

Detailed descriptions have been added to each table and image

Q8

Changes have been made to each of the reference descriptions

Q9

Changes have been made to all image and table formats

Specific comments

[Line 186] In section 2.2, the paper mentioned that CBAM modules were introduced into the ‘second layer of the generator’. Is it the second layer of the Single Scale generation or the second layer of the entire generator model?

It is the second layer of the Single Scale generation

[Line 189] I think Figure 2 should be the ‘overall single scale generation’, not the ‘Overall Architecture’. Please clarify what the architecture refers to.

It is indeed a ‘overall single scale generation’,which has been changed in this article

[Line 293] Does ‘sunken ships’ mean shipwreck? Please use the same term.

is a shipwreck, which has used the same term shipwreck

[Line 333] Figure 12 not found.

The language description is incorrect, it should be Figure 5, Figure 6, and it has been changed

[Line 346, 347] Where are the citations of Figure 5 and Figure 6?

The reference is Figure 12 of the previous question

[Line 355] 2 to 4 scales’ image lost. Also ‘sacales’ typo.

Changes have been made

[Line 395] What does the a,b,c,d,e modules mean in Figure 8?

ABCDE is the effect of amplification for different types of targets, and two more characteristic shipwreck are selected, and frogmen, planes, and schools of fish are selected

Author Response File: Author Response.pdf

Reviewer 2 Report

The article presents a sample augmentation method for sidescan sonar underwater target images based on CBL-sinGAN. While the article is well-written, there are some concerns that need to be addressed to improve the quality of publication.

The abstract of the article begins by addressing the scarcity and difficulty in acquiring sidescan sonar target images, which limits the application of deep learning algorithms in sidescan sonar target detection. However, mentioning the technique at an early stage is not suitable. The authors should first explain the research problem, aims, and objectives of the study, followed by existing solutions, research methodology, results, and conclusion with the significance of the research for the research community.

The introduction is well-written; however, the organization of the article should be presented at the end of the introduction section. Please ensure consistency by showing the full form first, followed by the abbreviation in brackets.

The heading "method" does not suit; please use a more suitable word.

In sub-heading 2.1, more gaps are observed. Additionally, in line 89, please provide a reference for the statement, and use equation numbers.

In line 333, "(See Figure 12)" the figure does not exist.

Regarding Figure 7, if there are no images for 2 scales, 3 scales, or 4 scales, please clarify.

It is observed that some results are presented in the discussion section. Either the section name can be renamed as "Experimental Results and Discussion," or the results can be moved into the result section, and only discussion should be provided here.

Future directions to carry out the work need to be included at the end of the conclusion section.

Most references are not recent and not in standard form.

Minor editing of English language required.

Author Response

Dear Editor and Reviewers,
Thanks very much for taking your time to review this manuscript. l really appreciate allyour comments and suggestions! Please find my itemized responses in below and myrevisions/corrections in the re-submitted files. Thanks again!

The article presents a sample augmentation method for sidescan sonar underwater target images based on CBL-sinGAN. While the article is well-written, there are some concerns that need to be addressed to improve the quality of publication.

The abstract of the article begins by addressing the scarcity and difficulty in acquiring sidescan sonar target images, which limits the application of deep learning algorithms in sidescan sonar target detection. However, mentioning the technique at an early stage is not suitable. The authors should first explain the research problem, aims, and objectives of the study, followed by existing solutions, research methodology, results, and conclusion with the significance of the research for the research community.

Re:The abstract section has been modified accordingly to include the research question, purpose and significance of the study

The introduction is well-written; however, the organization of the article should be presented at the end of the introduction section.

Re:The overall framework of this article has been added at the end of the introduction, and the corresponding organizational structure is explained

Please ensure consistency by showing the full form first, followed by the abbreviation in brackets.

Re:The corresponding wording has been revised. We think that the method is more suitable for this article, because the proposed CBL-sinGAN is itself a method, a means of amplification. The method is more suitable.

In sub-heading 2.1, more gaps are observed. Additionally, in line 89, please provide a reference for the statement, and use equation numbers.

Re:whats’ the mean ”gaps”?References have been provided for declarations and numbering has been used for each formula.

In line 333, "(See Figure 12)" the figure does not exist.

Regarding Figure 7, if there are no images for 2 scales, 3 scales, or 4 scales, please clarify.

Re:It has been modified and is actually Figure 5. In addition, the missing images of scales 2 and 3 have been supplemented.

It is observed that some results are presented in the discussion section. Either the section name can be renamed as "Experimental Results and Discussion," or the results can be moved into the result section, and only discussion should be provided here.

Re:This section has been revised to read: Experimental Results and Discussion.

Future directions to carry out the work need to be included at the end of the conclusion section.

Most references are not recent and not in standard form.

Re:A statement of the direction of future work has already been made. In addition, research articles and research results in the latest fields such as GAN adversarial neural network and side-scan sonar target automatic recognition have been added for citation, and the citation form has been unified to adopt the form of GB/T 7441.

Author Response File: Author Response.pdf

Reviewer 3 Report

This article proposes a sample augmentation method for side-scan Sonar underwater for target images based on CBL and sinGAN.

The authors should address the recommendations I have noted in their submitted article, mainly about improving their abstract and the introduction section, and add a survey of the related work section (Please refer to my comments on the document.)

Comments for author File: Comments.pdf

For the entirety of your document, please shorten your sentences.

Please revise the highlighted words and sections.

Author Response

Dear Editor and Reviewers,
Thanks very much for taking your time to review this manuscript. l really appreciate allyour comments and suggestions! Please find my itemized responses in below and myrevisions/corrections in the re-submitted files. Thanks again!

Comments：You need to add more context by adding a couple of general introduction sentences.

Re：We have added some concluding statements in the Abstract: to introduce the background and purpose of the paper

Comments：For the entirety of your document, please shorten your sentences.

Re：We've condensed some of the lengthy sentences in the article

Comments：These are more a description of the steps you took than a description of your method.

Re：Since the steps are innovative, they are highlighted in the abstract. We have also made appropriate revisions to the abstract to clarify the background and research objectives.

Comments：This is more of a related work section than an introduction.

An introduction is a wide entrance to the subject of your paper.

Re：In the introduction part, we do describe a lot of related work, but in the first three paragraphs, we also give a detailed introduction to the research status of related fields (side-scan sonar target recognition, side-scan sonar image amplification, and GAN adversarial neural networks).

Comments：What are your paper’s contributions.

Re：My paper’s contributions：Experiments in this paper prove that the single-image sample augmentation method for sidescan sonar underwater targets based on CBL-sinGAN proposed in this paper can generate a large number of high-quality sidescan sonar augmented samples according to the characteristics of sidescan sonar images, such as limited samples, monochromatic color, and diverse target shapes, Introducing a novel approach to address the issue of small target sample augmentation in sidescan sonar and high-performance underwater target detection models.

Comments：You need to make this more legible.

Re：We've enlarged the image and made it clearer

Comments：You need to center this image.

Re：Thanks for the suggestion, we've reformatted all of our images

Comments：Scales is misspelled in your figure.

Why are some of the boxes are empty?

Re：It could be a transfer issue or a WPS conversion issue, and we resubmitted it to make sure the image wasn't missing

Comments：How were these figure obtained?

Re：These are the FID, IS, and MMD indicators calculated by generating graphs at different scales

Comments：Future work?

Re：Future research directions have been added to the paper.

Author Response File: Author Response.pdf